Mexico’s UNAM weighs whether to rerun admissions exams amid AI-cheating concerns

Summer has turned into a torturous countdown for university students and faculty alike. Entrance exams at the National Autonomous University of Mexico (UNAM), the largest university in Mexico and Latin America, were held online and remotely for the first time and produced results so unusually strong that academic authorities have paused the admissions process to determine whether applicants used artificial intelligence to cheat.

The university now faces an unprecedented dilemma. A panel of 16 experts is expected to decide this week whether the scores should be validated or the exams administered again, with the start of the academic year fast approaching and new students due to enroll.

“This could set legal, technological, educational and evaluative precedents. It will teach us many lessons,” says Juan Manuel Flores Ayala, an expert in monitoring and evaluation in education. More than 158,000 applicants are affected, 22,000 places are at stake, and the reputation of the region’s benchmark university hangs in the balance.

The challenge facing the commission is a maze with no solution that will satisfy everyone. Students who secured places point to the hours of study and anxiety involved in getting through exams that define their future. But the cloud of doubt hanging over some of the successful candidates is forcing the university to take action.

“I would lean toward repeating the exam and moving the semester as needed to restore credibility and provide certainty to the people UNAM serves as a public institution,” says the expert. “There is a political and legal cost to doing so, but I think more would be gained than lost.”

Flores Ayala predicts a wave of complaints if the exams are ultimately retaken. If the university chooses instead to validate the results, he adds, it will need to explain clearly what went wrong and specify how it intends to deal with cases in which cheating is found to have occurred.

The challenge lies precisely in determining, after the fact and with surgical precision, which exams may have been compromised. Whether that can be done will largely determine whether the tests need to be administered again across the board, says Daniel Hernández Gutiérrez, coordinator of the degree program in Education and Digital Technologies at the Autonomous Metropolitan University (UAM).

If it is impossible to identify the tainted exams based on the data collected, he argues, “the tests would have to be repeated and, of course, not under the current conditions of an online exam, because that would simply recreate a process that was flawed from the outset.” All signs point to a return to in-person testing, with pen and paper.

Although time is running out and a decision cannot be delayed much longer, both experts believe there is still enough room to resolve the situation without damaging the university’s reputation. It would not be the first time the start of the academic year has been postponed, they note: UNAM has done so before because of strikes and walkouts, and the Covid-19 pandemic also forced adjustments to the academic calendar.

It was during that difficult period in 2020 that, out of necessity, universities began introducing the kind of online assessments that have since become a regular feature of university admissions and evaluation processes. However, experience with these systems remains limited, and institutions are still in the early stages of learning how to implement them effectively. Artificial intelligence is advancing at breakneck speed, while the safeguards needed to manage it do not always keep pace.

There are other experiences to draw on. The Autonomous Metropolitan University (UAM), where Hernández works, has also administered online, remote admissions exams in recent years without major problems.

“We have not had an extraordinary number of cases nor have we seen an increase in attempts to cheat or to use any AI tool,” the official says. “But at the same time, we can’t be too naive and assume the system can’t be circumvented.”

UAM uses a two-layer security system that requires applicants to use two devices: the computer on which they take the exam, equipped with monitoring software, and a second device, such as a tablet or smartphone, used for a simultaneous video call. According to Hernández, that approach ensures that oversight does not depend entirely on the effectiveness of the first mechanism.

At UNAM, by contrast, candidates were required to use only one device, monitored by artificial intelligence, while a human proctor remotely supervised every 150 students. It is still unclear whether the failure occurred in the monitoring system during the exam, whether the test questions were leaked, or whether the safeguards functioned as intended but were nevertheless bypassed by some candidates. Identifying where the breach occurred will be key to determining the most appropriate solution. “At least at a macro level, it is possible to know what happened,” Flores Ayala maintains.

There are also successful examples to look to, including Ticómetro, a diagnostic test developed by UNAM itself to assess students’ digital skills. The test involves a much smaller pool of candidates, about 35,000 students, and carries little consequence for participants’ academic futures. As a result, the incentives to cheat are greatly reduced, if not eliminated altogether.

The assessment was originally administered both online and in person until the pandemic forced it into a fully remote format. The shift did not alter the distribution of scores, explains Marina Kriscautzky, the former head of the department responsible for the test. It did, however, expose disparities in the technological resources available to students, particularly in low-income neighborhoods, where household noise and living conditions do not always meet the requirements of digital monitoring systems.

Two percent of UNAM’s exams have been canceled for “conduct contrary to what was established in the call.” In the absence of more details, it is not known whether any of those cases are part of the current controversy.

The episode now under public scrutiny brings together every conceivable challenge: the sheer scale of an exam taken by 158,000 applicants, placing enormous strain on monitoring systems; the high stakes for prospective students, whose academic futures can hinge on just a few hours; and the added complexity of remote testing, with candidates sitting the exam from home, each with their own circumstances.

UNAM has long demonstrated its ability to safeguard the integrity of admissions through traditional testing methods, a reputation suddenly called into question by the shift to a digital format. The challenge facing the expert panel is twofold: to find the least damaging solution for applicants who did not cheat on these crucial exams, and to preserve confidence in a process that serves as a benchmark for universities across Latin America.

Sign up for our weekly newsletter to get more English-language news coverage from EL PAÍS USA Edition